Intelligent Automation, Inc. — Department of Defense SBIR Phase I: N181-029

Description

There are needs to install automatic target recognition system on weapon system. In congested territory that includes enemy and friendly ships, it is essential for the missile to main target track with extremely low false alarm rate to minimize collateral damage. However, there are not many ATR systems transitions to weapon application as most developed ATR system are complex, computational expensive. In poor condition, where there is low return signal to high interference, ATR systems simply do not work.<br>To address this critical need, Intelligent Automation, Inc (IAI) proposes to develop an Intelligent Feature Extractor for Maritime Target (IFEMT) system based on deep learning. The key innovation of the proposed IFEMT system is the transfer-learning based approach adapting the existing deep convolutional neural networks (DCNN) e.g., GoogLeNet as feature extractor, a multi-class support vector machine (MC-SVM) classifier, and high fidelity physics-based synthetic ISAR data.